Chapter 22 Taxes on Savings

Introduction Chapter 22 Taxes on Savings Jonathan Gruber Public Finance and Public Policy Does the existing structure of income taxation in the United States reduce the amount of savings done by individuals? This is an important policy question because making more capital available to business can be a key determinant of economic growth. What is the appropriate role of capital income taxation the taxes levied on the returns from savings? aron S. elowitz - Copyright 2005 Worth Publishers Introduction This lesson proceeds as follows: Basic theoretical model with two-period consumption Empirical evidence Precautionary savings models Self-control models Retirement accounts TXTION ND SVINGS THEOR ND EVIDENCE The traditional theory of savings is to smooth consumption across periods. This is an implication of diminishing marginal utility of income. Intertemporal choice is the choice individuals make about how to allocate their consumption over time. s with hours of work in the labor supply model, savings is not valued directly, but is rather a means to an end. It can be thought of as a bad where the complementary good is future consumption. Figure 1 We define savings as the excess of current income over current consumption. It earns a real rate of return, r, that buys consumption in future periods. Figure 1 illustrates the basic model. (1+r) slope = -(1+r) slope = -(1+r(1-τ)) Initially savings is S, and consumption is. Taxing savings rotates the budget constraint, and creates income and substitution effects. S(1+r) B B S 1

The initial blue line represents the budget constraint, given income in period 1, B. The intertemporal budget constraint is the measure of the rate at which individuals can trade off consumption in one period for consumption in another period. Jack has preferences over consumption goods today and consumption in the future. He initially chooses bundle. The slope of the intertemporal budget constraint is (1+r), meaning that the opportunity cost of firstperiod consumption is the interest income not earned on savings for second period consumption. Savings (as with hours of work in the labor supply model), is measured going toward the origin on the x-axis. It is the difference between income and consumption. The model assumes the person can freely borrow, if his preferences dictate. Figure 2 If the government taxes all income, including interest income, then the rate of return falls from r to (1-t)r, because the government collects tr. The slope therefore changes from (1+r) to (1+((1- t)r), shifting the intertemporal budget constraint to red budget constraint, B. Figure 2 shows possible responses to the taxation of savings. * Substitution effect is larger Savings can fall. * Income effect is larger Or rise. B B B B C 1 * * S S The lower after-tax rate of return will cause an increase in first period consumption through the substitution effect. But the fall in the after-tax return makes Jack feel poorer, which reduces his consumption in the first period (and increases savings). The first panel shows that when the substitution effect dominates, savings falls. The second panel shows that when the income effect dominates, savings increases. One way to think about the income effect from a fall in the after-tax rate of return is to think about a target level of consumption in retirement. s the interest rate falls, Jack has to save more to meet this target level. 2

How does the after-tax interest rate affect savings? Unlike the empirical literature on labor supply, the empirical work on after-tax interest rates and savings has not reached a clear consensus. The elasticity of savings with respect to interest rates varies from 0 to 0.67. It is more difficult to compute the appropriate interest rate. In addition, it is more difficult to find appropriate treatment and control groups. Inflation and the taxation of savings In the late 1970s, the U.S. experienced double-digit inflation rates. t that time, neither the income tax brackets nor the treatment of capital income was indexed for inflation. This led to bracket creep whereby individuals would see an increase in their tax rate despite no real increase in their real income. lthough the treatment of income tax brackets changed in 1981 by becoming indexed to inflation, the rules about capital income taxation have remained the same. Inflation and the taxation of savings The nominal interest rate (i) is the interest rate earned by a given investment. The real interest rate (r) is the nominal interest rate minus the inflation rate. This measures an individual s actual improvement in purchasing power due to savings. These are related by r=i+π, where π is the inflation rate. The tax system taxes nominal returns, not real returns. Table 1 illustrates the impact of capital taxation in an inflationary environment. Table 1 Capital taxation in an inflationary environment Case Inflation Tax rate on interest Savings Nominal rate Interest earnings fter-tax resources Price of skittles Bags of skittles No inflation 0% 0% 100 $10 $110 $1.00 110 Inflation 0% 50% 0% 50% 100 100 100 $10 $10 $10 $105 $110 $105 $1.00 $1.10 $1.10 105 100 95.5 Constant real rate 0% 100 21% $21 $121 $1.10 110 50% 100 21% $21 $110.5 $1.10 100.5 The nominal rate will likely adjust for inflation, however. With taxes on nominal returns, the real return is negative! With inflation and a return, the real return is zero. Inflation and the taxation of savings In the first panel, there is no inflation. With no taxes, the number of skittles that can be bought is 110. With a 50% tax rate, only 105 bags can be bought. In the second panel, imagine that inflation equals the nominal rate of return. Thus, the real rate of return is 0%. With inflation but no taxes, 100 bags can be bought. With inflation and taxes, even though real purchasing power is unchanged, the taxation of nominal returns means only 95.5 bags can be bought. In the third panel, if the nominal rate adjusts for inflation (to 21%), then in the absence of taxation inflation will not erode the purchasing power of savings. Inflation and the taxation of savings The problem in the second and third panels, with taxation, is that taxes are levied on nominal, not real earnings. Individuals, when making savings decisions, care about the real interest rates. Because taxes are levied on nominal returns, the impacts of inflation on the tax code remain important. Higher inflation lowers the after-tax real return to savings. 3

LTERNTIVE MODELS OF SVINGS Precautionary saving models The precautionary saving model is a model of savings that accounts for the fact that individual savings serve at least partly to smooth consumption over future uncertainties. One of the most commonly given reasons for saving is for emergencies. This is a form of self-insurance. The intuition for precautionary savings are barriers to borrowing during an emergency. Liquidity constraints are barriers that limit the ability of individuals to borrow. Empirical Evidence Social insurance and personal savings There are a number of studies in support of the precautionary model. They show that greater uncertainty leads to higher savings, and that social insurance programs that lower income uncertainty lead to lower savings. Chou, et al. (2003) find that the introduction of National Health Insurance in Taiwan led to a decrease in savings among affected workers. Gruber and elowitz (1999) find that Medicaid expansions in the U.S. lowered the need for precautionary savings. lternative models of savings Self-control models n alternative formulation of the savings decision comes from behavioral economics models. Individuals have a long-run preference to ensure enough savings for smooth consumption throughout their lives, but their impatient short-run preferences may cause them to consume all their income and not save for future periods. These self control problems require commitment devices. lternative models of savings Self-control models Investment devices such as Christmas club bank accounts or tax-preferred retirement accounts are consistent with self-control problems. Self-control problems also may explain why individuals have substantial savings in illiquid forms (housing, retirement accounts), while at the same time carrying credit card balances at high interest rates. lternative models of savings Self-control models final piece of evidence for self-control problems is from an innovative experiment run by Thaler and Benartzi (2004), called Save More Tomorrow. Employees committed a portion of future pay increases to their retirement savings. By arranging the decision this way, the decision seemed less difficult. lthough this should not have any attractiveness to a rational saver, 78% of the employees offered the plan decided to join it, and 80% of those employees stuck with it through four pay raises. The bottom line is that Save More Tomorrow raised savings for employees. TX INCENTIVES FOR RETIREMENT SVINGS Because of concern about workers under-saving for retirement, the U.S. government has introduced a series of tax subsidies for retirement savings. There are four major incentives: Tax subsidy to employer-provided pensions 401(k) accounts Individual Retirement ccounts Keogh ccounts 4

vailable tax subsidies for retirement savings pension plan is an employer-sponsored plan through which employer and employees save on a (generally) tax-free basis for the employees retirement. defined benefit pension plan is one in which worker accrue pension rights during their tenure at the firm, and when they retire the firm pays them a benefit that is a function of the workers tenure at the firm and their earnings. defined contribution pension plan is one in which employers set aside a certain proportion of a worker s earnings in an investment account, and the worker receives this savings and any accumulated investment earnings when he or she retires. The percentage of earnings varies, but 5% is not uncommon. The contributions that employers make to pension plans are not taxed until the money is withdrawn at retirement. vailable tax subsidies for retirement savings 401(k) account is a tax-preferred retirement savings vehicle offered by employers, to which employers will often match employees contributions. The analog in the non-profit sector is 403(b). Many firms offer an employer match on contributions made to such a plan to encourage participation. The contribution limit for 2005 is $14,000, schedule to rise to $15,000 in 2006, and indexed for inflation thereafter. vailable tax subsidies for retirement savings n Individual Retirement ccount (IR) is a tax-favored retirement savings vehicle primarily for low-income and middle-income taxpayers, who make pretax contributions and are then taxed on future withdrawals. The contribution limit is currently $3,000 per person. vailable tax subsidies for retirement savings Keogh account is a retirement savings account specifically for the self-employed, under which up to $40,000 per year can be saved on a tax-free basis. They function much the same as 401(k) accounts, except they are not run through an employer. Why do tax subsidies raise the return to savings? ll of the tax subsidies have the following characteristics: Individuals avoid paying income tax on their contributions. Earnings accumulate at the before-tax rate of return. Withdrawals are taxed as ordinary income, not the lower capital gains tax rate. Why do tax subsidies raise the return to savings? Since taxes are paid at retirement, how are these accounts tax subsidized? The key ingredient is that you get to earn the interest on the money that would have otherwise been paid in taxes. This is composed of three important parts: The initial deductibility of the contributions Having earnings accumulate at the before-tax rate of return Having the potential to withdraw the money when a person is in a lower tax bracket. These tax subsidies can dramatically increase the rate of return to retirement savings. 5

Why do tax subsidies raise the return to savings? Table 2 illustrates the difference between a regular, taxed investment and a deductible IR. We assume in this table that the earnings accumulations are taxed as ordinary income, not as capital gains. Table 2 ccount type Earnings The tax advantage of IR savings Tax on earnings Initial deposit Interest earned (r=) Taxes paid upon withdrawal ( =25% Regular $100 ) $25 $75 $7.50 $1.88 =0.25x($7.50) IR $100 0 $100 $10 $27.50 =0.25x($110) With an IR (or 401k), the investment accrues at the before-tax rate of return. This tax subsidy leads to greater overall wealth. Total amount withdrawn $80.62 =$75+$7.50-$1.88 $82.50 =$100+$10-$27.50 These tax subsidies to retirement savings rotate the intertemporal budget constraint in an opposite way to the taxation. This is illustrated in Figure 3. 3 Figure 3 (1+r(1-τρ)) IRs increase the after-tax rate of return, and rotate the budget constraint. slope = -(1+r(1-τρ)) slope = -(1+r(1-τ)) B C S(1+r(1-τ)) B BC 3 = B The effect of IRs on savings is ambiguous, however. S Initially the after-tax rate of return is r(1-τ). With the tax-subsidy, however, the delay in tax payments reduces the tax burden by, so the effective tax rate on retirement savings is only τρ. Thus, the rate of return rises to r(1- τρ). The substitution effect leads to more savings, while the income effect leads to less savings. Thus, the change in total savings is ambiguous. Moving from point to B results in an increase in savings (where the substitution effect dominates), while moving from to C results in a decrease in savings (where the income effect dominates). One key institutional feature of 401(k) accounts, IRs, and so forth is that the annual contributions are capped. This creates a non-linearity in the budget constraint, where the tax-advantaged rate of return from saving below the cap is higher than taxed rate of return above the cap. Figure 4 illustrates this situation. 6

Figure 4 D slope = -(1+r(1-τ)) E With a cap, savings is subsidized, but only up to a point. The slope on segment BE is -(1+r(1- τρ)), while the slope on segment DE is -(1+r(1- τ)). This kinked budget constraint has potentially different effects on different types of savers. slope = -(1+r(1-τρ)) Consider an initially low saver in Figure 5a. B $3,000 Figure 5a Low saver B? C On the marginal $1 of savings, this individual now gets a higher rate of return. Thus, he faces both the income and substitution effects discussed before, so savings could increase (to point B) or decrease (to point C). On the other hand, consider an initially high saver in Figure 5b. Thus, the net effect is ambiguous for low savers. For a low saver, the income and substitution effects go in opposite directions. g 1,000 Figure 5b High saver For high-savers, IRs represent an income effect only and therefore lower savings. B C W 1 C W 2 $4,000 $5,000 On the marginal $1 of savings, this individual receives the same rate of return as before. He faces only an income effect, but not a substitution effect, so savings decreases (to point B). The income effect for high savers such as the person in panel (b) arises as they reshuffle their existing assets into an IR; they take $3,000 of savings they were already putting aside and label it as tax-preferred IR savings. It is possible that IRs actually lower overall private savings through this income effect. 7

pplication The Roth IR Roth IR is a variation on normal IRs to which taxpayers make after-tax contributions but may then make tax-free withdrawals later in life. Unlike traditional IRs, individuals are never required to make withdrawals from Roth IRs, so the earnings on assets can build up indefinitely. Why was the Roth IR created? Budget politics clearly came into play: traditional IRs entail immediate loss in tax revenue, while Roth IRs entail loss of future tax revenue. Implications of alternative models The retirement tax incentives may have stronger positive effects on savings than implied by the traditional theory and intertemporal budget constraint analysis. Consider, first, the precautionary savings motive. Imagine a person who had more than $3,000 in savings, but was using it for precaution against job loss. The illiquid IR would not be viewed as a good substitute, so contributions to the IR are not simply reshuffling. Thus, there may be more savings due to retirement incentives than is suggested by traditional models. Implications of alternative models Second, the hallmark of self-control models is the search for commitment devices to provide selfcontrol. 401(k) accounts, taken directly out of the paycheck, provide such a commitment device, because the money cannot be easily accessed until retirement. Beyond the demand for these accounts due to tax incentives will be demand that arises because of commitment. Private versus national savings The discussion so far has focused on private savings, but what matters for investment and growth is national savings. Retirement tax incentives have an offsetting effect on national savings because they are financed by a tax break. For example, imagine that 401(k)s raised private savings by 30 per $1 of contribution. If the tax rate were 43%, then 30 of tax revenue (43%x 70 of existing savings) is forgone, and there is no new national savings. Private versus national savings This example illustrates the notion of marginal impacts versus inframarginal impacts. The marginal impact is the 30 of new private saving. The inframarginal impact is the 70 of existing saving that was going to happen even in the absence of a tax subsidy. Empirical Evidence The impact of tax incentives for savings on savings behavior Empirical work examining the impact of IRs and 401(k)s has proven difficult. Those who contribute to IRs also save more in every form than non-savers who don t contribute to IRs. Thus, noncontributors do not form a good control group. The same kinds of comparisons have been made for workers in firms that offer a 401(k) to workers in firms that do not offer a 401(k). The workers in these different firms appear to be different in observable ways, and therefore are likely to be different in unobservable ways as well. 8

Empirical Evidence The impact of tax incentives for savings on savings behavior Engelhardt (1996) studies the Canadian Registered Home Ownership Savings Plan. Unlike the programs in the U.S., this created quasi-experimental variation because: It was available to renters but not owners. It was in effect from 1974 to 1985, but not afterwards. The benefit varied by marginal tax rate, which varied widely across Canadian provinces. The results indicated that each $1 contributed to the program resulted in up to 93 in new private savings, and up to 57 of new national savings. 9